Part 8 emphasizes building a supervised AI agent within a structured workflow using Temporal, focusing on retrieval, reasoning, and drafting without overspending on model tokens. The author outlines the importance of sequencing the development, integrating guardrails like budget limits, and asserts that trust in the agent comes from structured testing and supervision, not vague prompts.... Continue reading →

In Part 7 of the series, the author emphasizes the significance of durability in AI to minimize costs, particularly focusing on memoization and caching techniques. They outline strategies to prevent unnecessary token expenses during retries and propose a structured approach to optimize budget management before any billing occurs, ensuring efficient usage of resources throughout the... Continue reading →

Part 6 of 8: Durable AI for InterlinedList Every serverless AI feature I’ve built hits the same wall the second someone wants it at scale. “Can it tag all my untagged messages?” Sure, one message. “All of them?” Now we’re talking about a job that runs for minutes, touches hundreds of items, and will absolutely... Continue reading →

The post explores improvements to an AI tool for managing saved links, addressing challenges in retrieval. It proposes creating a durable cache for links and enhancing search functionality with vector embeddings. The author emphasizes the need for security in URL fetching and systematic database migrations. The goal is easy access to previously saved articles and... Continue reading →

The post discusses a critical bug in a scheduled publishing feature used for cross-posting on various platforms. It details how the current system may lead to double posts due to timing issues between cross-posting and database updates. The author suggests a robust solution involving a new scheduling system to avoid such errors while emphasizing the... Continue reading →